Speed test matrix
gpt-oss-120b on CMP 170HX 64GB ×2
vllm · 0.32.0rc1.dev249+g18305948a · MXFP45 contexts (0–120K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 120K
Hardware
- Type
- Discrete GPU
- GPU
- CMP 170HX
- GPU count
- 2
- VRAM
- 64 GB (2 × 32 GB)
- CPU
- 2x Intel Xeon E5-2630 v3
- RAM
- 64 GB
- OS
- Ubuntu 26.04.1
Engine
- Engine
- vllm
- Version
- 0.32.0rc1.dev249+g18305948a
- Quantization
- MXFP4
- Backend
- cuda
- Repository
- github.com/vllm-project/vllm
- Build
- ghcr.io/morrowmake/vllm-cmp170hx@sha256:5b45c4d6feafefe2eba962375b0a9a0b64fd22b9e05abb4d1966f14a32a2db79
Run
- Served model
- bench
- Revision
- main
- Benchmark tool
- llm-inference-bench v0.8.0
- Decode test
- Fixed duration
- Prefill mode
- Measured during decode warm-up
- Time per measurement
- 30 s
- Max output
- 8,192 tokens
Launch command
# Untuned out-of-the-box baseline. Spec decode: none. Layout: PP2 (stock vLLM flag) # 2x NVIDIA CMP 170HX 64GB (cmpunlocker), 180 W limit, PCIe Gen2 x4 | 2x Intel Xeon E5-2630 v3 | 64 GB DDR4 | Ubuntu 26.04.1, NVIDIA driver 610.57.04 | vLLM: Morrowmake vllm-cmp170hx image vllm serve openai/gpt-oss-120b --max-model-len 131072 --gpu-memory-utilization 0.94 --max-num-seqs 128 --enable-prefix-caching --trust-remote-code --pipeline-parallel-size 2
Prefill
prompt tok/s by context length8K context7,5481,074 ms TTFT
16K context6,7642,365 ms TTFT
32K context5,3965,891 ms TTFT
64K context3,72017,037 ms TTFT
120K context2,36450,190 ms TTFT
Sustained decode
total tok/s · rows are context length, columns are simultaneous requests| Context | 1 | 2 | 4 | 8 | 16 | 32 | 64 | 128 |
|---|---|---|---|---|---|---|---|---|
| 0 | 136.998 ms TTFT | 211.7119 ms TTFT | 351.6169 ms TTFT | 499.8183 ms TTFT | 617.9241 ms TTFT | 836.1329 ms TTFT | 1,343.2534 ms TTFT | 2,009.9840 ms TTFT |
| 16K | 113.8155 ms TTFT | 190.5215 ms TTFT | 308.7305 ms TTFT | 434.0523 ms TTFT | 579.9934 ms TTFT | 628.01,557 ms TTFT | Over capacity | Over capacity |
| 32K | 94.0222 ms TTFT | 168.7345 ms TTFT | 249.7529 ms TTFT | 351.5871 ms TTFT | 490.71,379 ms TTFT | 541.72,442 ms TTFT | Over capacity | Over capacity |
| 64K | 72.0356 ms TTFT | 118.7577 ms TTFT | 199.8707 ms TTFT | 235.81,385 ms TTFT | 347.92,548 ms TTFT | Over capacity | Over capacity | Over capacity |
| 120K | 50.6521 ms TTFT | 93.5869 ms TTFT | 143.91,156 ms TTFT | 186.62,388 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity |
Over capacity: not run because this many requests at this context would exceed the server's KV cache.
TTFT is measured after the benchmark warms the server's prompt cache, so it mostly excludes prefill.
All measurements
Latencies are medians. Total tok/s is output across all simultaneous requests.| Phase | Context | Requests at once | Status | Total tok/s | Per request tok/s | TTFT | Time between tokens | Average active requests | Completed requests | Output tokens | Measured for |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Prefill | 8K | 1 | Valid | 7,548.0 | — | 1,074 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 6,764.0 | — | 2,365 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 5,396.0 | — | 5,891 ms | — | — | 1 | — | — |
| Prefill | 64K | 1 | Valid | 3,720.0 | — | 17,037 ms | — | — | 1 | — | — |
| Prefill | 120K | 1 | Valid | 2,364.0 | — | 50,190 ms | — | — | 1 | — | — |
| Sustained decode | 0 | 1 | Valid | 136.9 | 138.6 | 98 ms | 7.2 ms | 1.0 | 0 | 1,141 | 8.3 s |
| Sustained decode | 0 | 2 | Valid | 211.7 | 106.9 | 119 ms | 9.4 ms | 2.0 | 0 | 6,351 | 30.0 s |
| Sustained decode | 0 | 4 | Valid | 351.6 | 88.0 | 169 ms | 11.4 ms | 4.0 | 0 | 10,548 | 30.0 s |
| Sustained decode | 0 | 8 | Valid | 499.8 | 63.8 | 183 ms | 15.7 ms | 8.0 | 0 | 15,008 | 30.0 s |
| Sustained decode | 0 | 16 | Valid | 617.9 | 42.9 | 241 ms | 23.3 ms | 16.0 | 0 | 18,548 | 30.0 s |
| Sustained decode | 0 | 32 | Valid | 836.1 | 32.9 | 329 ms | 30.4 ms | 32.0 | 0 | 25,084 | 30.0 s |
| Sustained decode | 0 | 64 | Valid | 1,343.2 | 24.9 | 534 ms | 40.2 ms | 64.0 | 0 | 40,262 | 30.0 s |
| Sustained decode | 0 | 128 | Valid | 2,009.9 | 18.8 | 840 ms | 53.1 ms | 128.0 | 0 | 60,272 | 30.0 s |
| Sustained decode | 16K | 1 | Valid | 113.8 | 109.9 | 155 ms | 9.1 ms | 1.0 | 0 | 3,413 | 30.0 s |
| Sustained decode | 16K | 2 | Valid | 190.5 | 95.5 | 215 ms | 10.5 ms | 2.0 | 0 | 5,714 | 30.0 s |
| Sustained decode | 16K | 4 | Valid | 308.7 | 76.3 | 305 ms | 13.1 ms | 4.0 | 0 | 9,261 | 30.0 s |
| Sustained decode | 16K | 8 | Valid | 434.0 | 55.4 | 523 ms | 18.1 ms | 8.0 | 0 | 13,018 | 30.0 s |
| Sustained decode | 16K | 16 | Valid | 579.9 | 37.4 | 934 ms | 26.8 ms | 16.0 | 0 | 17,397 | 30.0 s |
| Sustained decode | 16K | 32 | Valid | 628.0 | 24.3 | 1,557 ms | 41.1 ms | 32.0 | 0 | 18,832 | 30.0 s |
| Sustained decode | 16K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 1 | Valid | 94.0 | 93.2 | 222 ms | 10.7 ms | 1.0 | 0 | 2,813 | 29.9 s |
| Sustained decode | 32K | 2 | Valid | 168.7 | 82.4 | 345 ms | 12.1 ms | 2.0 | 0 | 5,060 | 30.0 s |
| Sustained decode | 32K | 4 | Valid | 249.7 | 62.8 | 529 ms | 15.9 ms | 4.0 | 0 | 7,491 | 30.0 s |
| Sustained decode | 32K | 8 | Valid | 351.5 | 44.7 | 871 ms | 22.4 ms | 8.0 | 0 | 10,539 | 30.0 s |
| Sustained decode | 32K | 16 | Valid | 490.7 | 31.6 | 1,379 ms | 31.6 ms | 16.0 | 0 | 14,709 | 30.0 s |
| Sustained decode | 32K | 32 | Valid | 541.7 | 18.5 | 2,442 ms | 54.1 ms | 32.0 | 0 | 16,244 | 30.0 s |
| Sustained decode | 32K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 64K | 1 | Valid | 72.0 | 71.6 | 356 ms | 14.0 ms | 1.0 | 0 | 2,160 | 30.0 s |
| Sustained decode | 64K | 2 | Valid | 118.7 | 59.6 | 577 ms | 16.8 ms | 2.0 | 0 | 3,559 | 30.0 s |
| Sustained decode | 64K | 4 | Valid | 199.8 | 50.2 | 707 ms | 19.9 ms | 4.0 | 0 | 5,993 | 30.0 s |
| Sustained decode | 64K | 8 | Valid | 235.8 | 33.4 | 1,385 ms | 29.9 ms | 8.0 | 0 | 7,073 | 30.0 s |
| Sustained decode | 64K | 16 | Valid | 347.9 | 23.5 | 2,548 ms | 42.5 ms | 16.0 | 0 | 10,433 | 30.0 s |
| Sustained decode | 64K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 64K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 64K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 120K | 1 | Valid | 50.6 | 50.8 | 521 ms | 19.7 ms | 1.0 | 0 | 1,519 | 30.0 s |
| Sustained decode | 120K | 2 | Valid | 93.5 | 47.2 | 869 ms | 21.2 ms | 2.0 | 0 | 2,806 | 30.0 s |
| Sustained decode | 120K | 4 | Valid | 143.9 | 36.2 | 1,156 ms | 27.6 ms | 4.0 | 0 | 4,316 | 30.0 s |
| Sustained decode | 120K | 8 | Valid | 186.6 | 23.6 | 2,388 ms | 42.4 ms | 8.0 | 0 | 5,596 | 30.0 s |
| Sustained decode | 120K | 16 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 120K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 120K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 120K | 128 | Over capacity | — | — | — | — | — | — | — | — |
Over capacity: not run because this many requests at this context would exceed the server's KV cache.